2 min Applications

EU users can keep conversations with Google Bard out of training set

Chatbot Bard has a Privacy Hub in Europe. That allows users to keep their conversations with Bard out of the training set of machine learning technologies.

Google released his chatbot Bard to the European public yesterday. TechCrunch then sought to find out why the chatbot is now allowed to come to the European public when it was previously blocked by Ireland’s Data Protection Commission (DPC).

“Google made a number of changes prior to the launch of [Bard], notably increased transparency and changes to user controls. We will continue our collaboration with Google regarding Bard post-launch and Google has agreed to conduct a review and provide a report to the DPC once Bard has been operational in the EU for three months,” replied Graham Doyle, media and communications manager at the DPC.

Privacy Hub

A spokesman at Google points mainly in the direction of the Bard Privacy Hub in response. “Users can also choose how long Bard keeps their data on their Google account. By default, Google keeps Bard activity in the user’s Google account for up to 18 months, but users can change this to three or 36 months if they wish. They can also turn this off completely and easily delete their Bard activity at g.co/bard/myactivity.”

Google collects by default Bard’s conversations with all users. The company does this to improve existing products and develop new ones. The company says it keeps all data for a maximum of three years.

However, the Privacy Hub allows users to keep their conversations with Bard out of Google’s hands. New conversations with the chatbot then no longer enter the systems to train machine learning models. Only if the user provides feedback on a response from Bard or marks the response as harmful will the tech giant still gain access to the conversation.

Users who notice that the chatbot provides wrong information about them can further try to change that as well. It is possible to ask to remove content from the chatbot via a request.

Also read: European websites do not appear to train Google’s AI models